ETC5521 Assignment 2

Australian Animal Outcomes

This assignment is for ETC5521 Assignment 2 by Team galah comprising of (Sen Wang), (Muhammad Soban Qasim), Lachlan Moody, and Xitong He.

1 Introduction and motivation

The RSPCA is Australia’s oldest, largest and most trusted animal welfare organization and, according to its description, there is a gradual increase in the number of animals received each year in animal shelters and adoption centers across the country (“About Us,” n.d.). With this comes an ever increasing amount of pressure and stress on the organisation as it is forced to develop an abundance of new methods to deal with this growing count.

In addition to this, many articles and posts online have recently been indirectly blaming the RSPCA for their high euthanization rates in their centers as some see the worst side of the organisation Powell (2020). This thus motivated the following primary question for the subsequent analysis:

How does the RSPCA manage animal outcomes across Australia?

Once this research area was established, 6 subsequent analysis questions were determined. They were:

1. How effective are the RSPCA at producing ‘good outcomes’ for animals in Australia?
   
2. Is there a relationship between the type of animal and its outcome?
   
3. Are cats more popular as pets than dogs?
4. What are the most frequent complaints in Brisbane?
5. Who gets the most complaints between dogs and cats?
   
6. How are the different type of complaints are distributed in dogs and cats?

During the research conducted, multiple articles and web pages dedicated to the adoption of animals that have been in the care of RSPCA were uncovered. If you are interested in adopting an animal, please click this link. (“Adopt a Pet,” n.d.)

2 Data description

In this section, we will mainly introduce the data structure, data sources and data description.

There are three data sets used on this analysis, and the cleaned data is obtained from GitHub tidytuesday.

The animal_outcome data source is from source website, which is the annual data from 1999 to 2018, consisting of 664 observation. It contains eight main states in Australia and the data structure and detail is shown as the below tables.

Table 2.1: Animal outcomes data

Variable	Class	Description
year	double	Full year
animal_type	character	Animal type (horse, wildlife, dog, cat, etc)
outcome	character	Animal outcome - euthanized, released, rehomed, etc
ACT	double	ACT - Australian Capital Territory
NSW	double	New South Wales
NT	double	Northern Territory
QLD	double	Queensland
SA	double	South Australia
TAS	double	Tasmania
VIC	double	Victoria
WA	double	Western Australian
Total	double	Australian Total

The animal_complaints data is from Australia government data, and the cleaned data is obtained from Github tidytuesday which is from April 2014 to May 2020. It also contain 42413 observation with different type of animals and complaints. The variable information included in the data is displayed in the following table.

Table 2.2: Animal complaints data

Variable	Class	Description
Animal Type	character	Animal Type
Complaint Type	character	Complaint type
Date Received	character	Date received (Month - year)
Suburb	character	Suburb/region
Electoral Division	character	Electoral Division

The brisbane_complaints data focus on the Brisbane city and it comes form Brisbane city council, and the cleaned data is from Github tidytuesday. 31330 observation are recorded from February 2016 to December 2019 to in this dataset, and the animal type and complaint type are mainly used in this analysis. The following table is shown more details in this dataset:

Table 2.3: Animal complaints data

Variable	Class	Description
nature	character	Nature of complaints (animal)
animal_type	character	Animal type
category	character	Category of complaint
suburb	character	Suburb where reported
date_range	character	Date range (typically 1 quarter + year)
responsible_office	character	Responsible office for the complaint
city	character	City (Brisbane)

2.1 Limitations

The animal_outcomes data set has not specified that suburbs, which limit our approach to identifying which suburb has the most common complaint type. This can also help to identify which suburb has the most animal type too by looking into the animal type the complaint was made against. Moreover, the number for the animals outcome might be effected by the different legislation governing the animal welfare.

For the animal_complaints data ,the Suburb variable is only recorded the suburb from Queensland which makes analyze complaint issue more specific and not in the general perspective. Another limitation in this dataset is about animal type which are only contained two animals such as cat and dogs so it is limited to analyze people complaints problems and raise the deviation in results.

In the brisbane_complaints data set, the date variable range is difficult to apprehend, because of fluctuating ranges and inconsistencies. In addition, it is hard to re-defined the date variable as it mixed different types of date range such as quarterly and monthly data. Furthermore, Brisbane is the capital of Queensland, indicating that this dataset is still under Queensland State but the suburb variable become to be specified in the Brisbane city.

3 Questions

3.1 Q1: How effective are the RSPCA at producing ‘good outcomes’ for animals in Australia?

The primary goal of the RSPCA, as stated on their website, is the contribution to “animal welfare” in Australia. Many animals are taken into care by the RSPCA that are neglected, reported, lost or require professional care. However, sometimes the RSPCA are forced into make the tough decision to euthanize animals when it is the most humane alternative. This could be due to causes such as medical issues or behavioral problems.

So how effective are the RSPCA at achieving their goal? This question will analyse the RSPCA’s success at achieving ‘good outcomes’ for the animals that come into their care.

For this analysis, the outcomes; rehomed, in care, in stock, reclaimed, released, transferred and currently in care were considered as ‘good outcomes’, essentially anything beside euthanization. Let us look at the animals saved by RSPCA or the rescue rate.

Figure 3.1: Positive animal outcomes by state

Figure 3.1, shows the rate of animals that were in the care of the RSPCA that experienced a ‘good outcome’, from the year 1999 to 2018 in each state. The plot clearly shows that Western Australia outperformed all other Australian states, recording a rate of over 75% for positive outcomes. In comparison, New South Wales had the lowest ‘good outcome’ rate at just 53%, followed closely by Queensland at 53%. This suggests that almost half of all animals received by the RSPCA over this time in these states were euthanized, well behind the performance of the other states. Thus, these areas should be of focus for the RSPCA. As this graphic was made interactive, users can hover over any state to explore the exact rate in each region.

So while it definitely appears that certain states are outperforming others, it is important to consider how this rate has changed over time. Before comparing statistics on a state-by-state basis however, a national overview was produced below in Figure 3.2. This graphic displays density plots of the distribution of positive outcomes for each year contained within the data set. Reading down the years, a clear improvement in performance can be seen. Beginning in 1999, not only is the average value lower, there is also a lot of variability in the rates recorded between states. However, over the years up until 2018, the average rate improves to approximately 75% nationally without the multi-modal distribution observed previously. This suggests that the RSPCA’s success at achieving their primary goal has indeed gotten better over the years.

Figure 3.2: Change in positive outcomes over time, with Year on the y-axis and the percentage of positive outcomes on the x-axis

Following this, the outcomes over time were further broken down into a state-by-state basis for comparison with Figure 3.1. Figure 3.3 paints a more positive picture than the map examined earlier. Every state in the country has shown a general trend of improvement over the years captured by the data. New South Wales in particular, the lowest overall, has shown the greatest improvement in recent years. Mousing over the last recorded year, it can be seen that over 96% of animals experienced a ‘good outcome’ from the RSPCA in 2020. This is far above their overall average of 53%. The only state not too show obvious improvement was Western Australia. The highest averaging state instead maintained its rate over the time frame, with a noticeable drop in 2008 to 52%. This is quite unusual considering its overall performance and the fact that rates of 91% were recorded in both 2007 and 2009. Despite this, the trend across the country is positive for both the RSPCA and the future of Australian animals.

Figure 3.3: Change in positve outcomes over time per state with percentage of positive outcomes on the x-axis and the year on the y-axis, faceted by state

Speaking of Australian animals, a final measure to evaluate the RSPCA on is their performance by animal over time.This relationship is explored in Figure 3.4. Positively, there is some improvement seen in the outcome of Dogs, Cats and Other Animals in particular with all showing an increased rate of positive outcomes over time. Whereas Horses, Livestock and Wildlife have not shown any observable improvement. Of particular concern is the Wildlife category. While Horses and Livestock have maintained reasonably high rates (above 75%), positive outcomes for Wildlife have hovered around 50%. This thus suggests that almost all wildlife received by the RSPCA are being put down. This is not ideal and thus should be area of improvement for the organistaion to produce better outcomes for these animals.

Figure 3.4: Change in positve outcomes over time by animal with rate on the y-axis and year on the x-axis, faceted by animal type

3.2 Q2: Is there a relationship between the type of animal and its outcome?

Expanding on the observations made above in Figure 3.4, a second area of interest was considered relating to the relationship between the different types of animal and any one individual outcome. This was chosen as potentially troubling outcomes associated with any one animal could be uncovered and highlighted, leading to better intervention measures by the RSPCA. Figure 3.5 below depicts the different proportionate outcomes for each animal across the different outcome types. A proportion was chosen for more direct comparison as the quantity received by the RSPCA was not consistent across the various animal types.
There are several observations that can be made from the plot. Firstly, the most euthanized animals were Wildlife and Cats by quite a significant margin. This suggests that there are fewer options available when it comes to these two animals. Cats in particular appear to only be euthanized or rehomed with very few owners ever reclaiming their animal. Whereas Dogs for example are much more evenly distributed between euthanized, reclaimed and rehomed indicating a higher change of the dog returning to its original owner. Wildlife meanwhile appears to have its own special category of released, which is reasonable considering most domesticated animals wouldn’t fare well in the wild. Interestingly, Horses clearly have the highest rate for both currently in care and in stock. This may highlight a willingness to hold onto these animals for a longer period of time compared to others, with most being rehomed at some point.

Figure 3.5: Proportionate outcomes for different animals with proportionate outcome on the y-axis and the animal type on the x-axis, faceted by outcome

As was done before, this measure can also be examined over time. Figure 3.6 shows this relationship for each outcome and animal type. Examining this plot, several key observations can be made. Firstly, the proportion of wildlife released decreased quite significantly in the early 2000’s in favor of animal transfers. This may indicate a shift in the strategy of the organisation. Secondly, the proportion of all animals being euthanized had decreased in recent years, particularly for Cats which is especially positive considering the results above. Wildlife however, did not follow this trend, dipping down to 30% in 2012 before returning to its normal level of around 60% the next year. Perhaps the strategies used during this year should be reconsidered by the RSPCA. Thirdly, the rate of Dogs being reclaimed has greatly increased, almost doubling from 22% in 1999 to 39% in 2018. Thus, the strategies used for Dogs may provide a template for the organisation to use in its care of other animals. Finally, the level of animals in stock and currently in care has been increasing for horses over time but not for any other animal. This anomaly should be further investigated (note that ‘Currently In Care’ appears to have replaced ‘In Stock’ as a classification in 2009).

Figure 3.6: Change in proportionate animal outcomes over time with proportionate outcome on the y-axis and year on the x-axis, faceted by outcome and coloured by animal type

Two further important factors to consider is if these animal outcomes are different for each state and how this has changed over time. To examine this, an animation was made and is displayed in Figure 3.7 below. Each state is highlighted based on the most common proportional outcome for each animal over each year in the data set. Aside from there being some apparent missing data, there are several important findings highlighted by this plot. Firstly, beginning with Cats, it is positive to see that euthanization became less and less common across the country over time in favor of rehoming. Similarly, by 2011, in no Australian state was euthanization the most common outcome for dogs. Horses meanwhile displayed much greater variability in the most common outcome by state indicating perhaps a lack of trend or plan for this animal type. Livestock and Other Animals meanwhile displayed similar trends to Cats with rehoming becoming the preferred method over euthanization. Finally, wildlife appears to have the most missing data with euthanized being the most commonly seen outcome over the time frame with a trend towards released and transferred in the Northern Territory and South Australia in particular in more recent years. As a final note, Western Australia rarely has euthanization as its most common outcome for any animal type which is supports the findings made in Figure 3.1.

Figure 3.7: Change in proportionate animal outcomes over time by state for each animal type

Speaking of euthanization, Figure 3.8, shows us the euthanization count by animal type throughout Australia from the year 1999 to 2018. Cats have the highest recorded euthanization count at 553,956 in Australia under the care of RSPCA. So is it fair to say that Cats are more likely to euthanized than other animals? This question can be answered with modeling.

Figure 3.8: Animal Euthanization count

The table below attempts to capture the linear relationship between outcome and animal type. The interactivity of the table allows the exploration of several different variables. Firstly, sorting by the r squared value, which measures the explanatory power of the model, it can be seen that the outcome animals being in care by the RSPCA is most directly tied to animal type, followed closely by reclaimed. The output can be restricted to just the ‘InCare’ model by typing this in the box in the top right. Sorting then by the estimate term, it can be seen that Horses are most positively associated with this outcome whereas Wildlife are the lowest. Noticeably the difference between Horses and second highest is much larger than Wildlife and the next lowest. Removing the filter and again sorting by estimate, it can be seen that Dogs actually have the highest positive association with any one outcome - that being reclaimed. Similarly, if the sort is reversed it can be seen that Horses are most negatively associated with euthanization rate followed closely by Dogs. Notably for the relationships above all p values were below the 0.05 significance level indicating that the animal type was significant on the outcome. As the table is interactive, all formed models can be explored.

3.3 Q3: Are cats more popuplar as pets than dogs?

Were the dogs or cats more likely to get reclaimed or rehomed in Australia? Were dogs popular than cats? Companion animals are an important part of our social world. We often talk to them as if they were humans and some even refer to pets as their children.

In this section, we will explore the outcome distribution between the two animals and emphasize in the reclaimed and rehomed outcome, relating to which kind of animals are getting popular in human life.

Figure 3.9: Outcome distribution between dogs and cats

Figure 3.9 shows how the different outcomes have distributed over time in Australia for dogs and cats based RSPCA data. There are seven outcomes contained in these two animals and the euthanized outcome is more than 50% in cats, whereas is only around 31% in dogs. Of the dog that were euthanased, the majority of cases are because of severe behavioral issues such as aggression and severe anxiety (Bartlett et al. 2005). However,the currently in care for cats is 3.05% more than dogs,revealing the life of cats are shorter and are easier to get disease than dogs in general so that the euthanized outcome is higher than dogs. In addition, the combined with reclaimed and rehomed outcome are 40.7% in cats, but there are more than half in dogs. In fact, the euthanized, reclaimed and rehomed outcome in dogs are distributed at 30% on average,indicating that dogs get reclaimed and rehomed outcome more than euthanized,indicating that people are more likely to reclaimed or rehomed dogs instead of cats. In addition, the inventory of cats is still 1.16%, while the inventory of dogs is only 0.766%, further indicating that dogs are more popular than cats for adopting program. More people reported liking dogs than reported liking cats, a finding consistent with stereotypes about the friendliness of dogs and the aloofness of cats. And having a dog is contributed to cardiovascular health benefits, both in terms of length of survival (Somervill et al. 2008).

In general, there is a general belief that dogs are more popular pets than cats based on RSPCA data.

3.4 Q4: What are the most frequent complaints in Brisbane?

By analyzing more specific animal complaints problem, we focus on Brisbane city which in the capital of Queensland state. In this section, we will explore the distribution with different complaints and the most frequent complaints in Brisbane.

Figure 3.10: Different types of complaint in Brisbane

From Figure 3.10, shows us that the distribution of different complaints with animals in the Brisbane. The highest proportion of animal complaints is fencing issue, followed by wandering animals, attack on animals, and attack on a person. The reason about this four categories are happened in most is due to people didn’t build up good condition in fences or enclosure for the animals which are easier for them to escape out of the fence and wander in the street. The animals attacks are often caused by animals wandering in the street or rushing out from poorly fenced properties(Hindle 1992). Moreover, once they run out from the fence or enclosure and wander in the street, it would increase the probability for them to attack people as some animals are aggressive and out-of-control in unknown situation.

3.5 Q5: Who gets the most complaints between dogs and cats?

Figure 3.11: The total number of complaints in dogs and cats

Figure 3.11, shows that dog complaints made by people are disproportionately higher than complaints against cats across different years. Thus, the total number of complaints in dogs are more than cats.

3.6 Q6: How the different type of complaints are distributed in dogs and cats acorss time?

Instead of focusing on which kind of animals gets the most of complaints, we are more interested in about what type of complaints have higher probability to happened between dogs and cats? And this new question is extended the the scope of the original analysis with animal complaints which only simply considered about the total number of complaints between dogs and cats. Thus, in this section,the distributions between dogs and cats with different type of complaints across several years are analyzed in more details.

Table 3.1: The total number of different complaint types in dogs and cats

Animal Type	Complaint Type	Total	percentage
cat	Wandering	438	10.7%
cat	Enclosure	610	14.9%
cat	Private Impound	3046	74.4%
dog	Attack	3836	10.01%
dog	Aggressive Animal	4459	11.64%
dog	Wandering	5501	14.36%
dog	Enclosure	5672	14.8%
dog	Noise	9205	24.02%
dog	Private Impound	9646	25.17%

Based on the Table 3.1, cats has three common complaints such as Enclosure,Private Impound and Wandering, but for dogs, beside of these three, they are also complained by aggressive animals, attack and noise. Compared with the wandering complaints, cats have 438 case whereas the dogs have 5501 case,almost 12 times as many as cats,suggesting that dogs get more complaints in wandering category. Similarly,by comparing the enclosure complaints, the case happened in dogs are almost 9 times in cats.

However, it is not an appropriate method to explain the complaints issue between cats and dogs in numerical comparison as their population are different. For example, the private impound is the most of case as one of complaints in dogs. In fact, private impound complaints have more than 74% in cats but dogs only have around 25% in all complaints. In order to compare with different type of complaints across time between cats and dogs, we selected same type of complaints attribute in two animals to construct mosaic plot which can be visualized directly about the complaints distribution between cats and dogs.

Figure 3.12: Animal Compplaints distribution across time

From the Figure 3.12, it shows that how the three common types of complaint distributed across several years between cats and dogs. There is an increasing trend with higher proportion in wander complaint for dogs,indicating that dogs get more complaints with wandering complaint than cats in the recent years. Dogs were more likely to be reported as having been abandoned in the recent ten years which would increase the probability in wandering complaint for dogs(Shih, Paterson, and Phillips 2019). Moreover, based on the Table 3.1, dogs complaints with aggressive animal and attack issue are around 22%,further explaining that it can increase security risks, such as attacks from other dogs, or attacks on humans while dogs wandering on the streets so makes more people complain wander issue. In addition,dogs have higher proportion than cats in enclosure complaint with increasing trend in recent years as some dogs lives in poor condition enclosure like offering insufficient food and water which have apparently became more prevalent(Beaver 1994). On the other hands, for the private impound, cats are most likely to get complaints more than dogs. One of the reasons is that cats are easy to be impounded privately, once they are found on the street without identifying their owners, and they are small-sized animals as well. Therefore, private impound complaints have most probability to happen in cats whereas enclosure and wander complaints are reported in dogs.

4 Conclusion

This report has analysed several different ways in which the RSPCA managed animal outcomes across Australia - the original question of interest for the report. The key findings can be summarised as such: Firstly, Western Australia is the strongest performing state in producing positive outcomes for animals and also this rate has increased over time for all states showing an improved performance for the organisation. Efforts however relating to wildlife in this area are lagging behind the other animal types. Secondly there does appear to be some correlation between animal type and its associated outcome. This was evidenced by the observed proportional outcomes for each animal type in the data set and was modeled. Thirdly, based on RSPCA data dogs were found to be the most popular animal, more than cats even. Fourthly, the most common complaints in Brisbane were related to fencing issues followed by wandering animals and attacks on both other animals and people. It was hypothesised that these three complaints were related due to animals escaping their enclosure and being in an unfamiliar environment. Additionally, the reports received for dogs was exponentially higher than that for cats. Finally, the complaints received for both cats and dogs were proportionally different, indicating that these issues may be related to the animal specifically. This finding, and those listed above, may help the RSPCA in designing more effective management strategies for the growing number of animals taken into their care.

5 Acknowlegments

The R packages we used in this report: Hughes (2020) Wickham and Henry (2020),
Wickham (2016), Wickham et al. (2020), Bivand, Keitt, and Rowlingson (2020), Bivand and Lewin-Koh (2020), Wickham (2011), Bivand and Rundel (2020), Sievert (2020), R by Ray Brownrigg, Minka, and Plan 9 codebase by Roger Bivand. (2020), Cheng, Karambelkar, and Xie (2019), Tennekes (2020), Pebesma (2018), Robinson, Hayes, and Couch (2020), Zhu (2019), Wilkins (2019)

References

“About Us.” n.d. RSPCA Australia. https://www.rspca.org.au/about-us.

“Adopt a Pet.” n.d. RSPCA Australia. https://www.rspca.org.au/adopt-pet.

Bartlett, Paul C, Andrew Bartlett, Sally Walshaw, and Stephen Halstead. 2005. “Rates of Euthanasia and Adoption for Dogs and Cats in Michigan Animal Shelters.” Journal of Applied Animal Welfare Science 8 (2): 97–104.

Beaver, Bonnie V. 1994. “Owner Complaints About Canine Behavior.” Journal of the American Veterinary Medical Association 204 (12): 1953–5.

Bivand, Roger, Tim Keitt, and Barry Rowlingson. 2020. Rgdal: Bindings for the ’Geospatial’ Data Abstraction Library. https://CRAN.R-project.org/package=rgdal.

Bivand, Roger, and Nicholas Lewin-Koh. 2020. Maptools: Tools for Handling Spatial Objects. https://CRAN.R-project.org/package=maptools.

Bivand, Roger, and Colin Rundel. 2020. Rgeos: Interface to Geometry Engine - Open Source (’Geos’). https://CRAN.R-project.org/package=rgeos.

Cheng, Joe, Bhaskar Karambelkar, and Yihui Xie. 2019. Leaflet: Create Interactive Web Maps with the Javascript ’Leaflet’ Library. https://CRAN.R-project.org/package=leaflet.

Hindle, Anne. 1992. “Designing Community Education Programs to Promote Animal Welfare: The Rspca’s Experience.” In Proceedings of the Urban Animal Management Conference., Mackay. Citeseer.

Hughes, Ellis. 2020. TidytuesdayR: Access the Weekly ’Tidytuesday’ Project Dataset. https://CRAN.R-project.org/package=tidytuesdayR.

Pebesma, Edzer. 2018. “Simple Features for R: Standardized Support for Spatial Vector Data.” The R Journal 10 (1): 439–46. https://doi.org/10.32614/RJ-2018-009.

Powell, Rebeka. 2020. “RSPCA ’Sorry’ After Euthanasing 11-Week-Old Kitten Without Consent.” ABC News. ABC News. https://mobile.abc.net.au/news/2020-02-26/rscpa-qld-issues-apology-after-accidentally-euthanasing-kitten/12001668.

R by Ray Brownrigg, Doug McIlroy. Packaged for, Thomas P Minka, and transition to Plan 9 codebase by Roger Bivand. 2020. Mapproj: Map Projections. https://CRAN.R-project.org/package=mapproj.

Robinson, David, Alex Hayes, and Simon Couch. 2020. Broom: Convert Statistical Objects into Tidy Tibbles. https://CRAN.R-project.org/package=broom.

Shih, Hao Yu, Mandy Paterson, and Clive JC Phillips. 2019. “A Retrospective Analysis of Complaints to Rspca Queensland, Australia, About Dog Welfare.” Animals 9 (5): 282.

Sievert, Carson. 2020. Interactive Web-Based Data Visualization with R, Plotly, and Shiny. Chapman; Hall/CRC. https://plotly-r.com.

Somervill, John W, Yana A Kruglikova, Renee L Robertson, Leta M Hanson, and Otto H MacLin. 2008. “Physiological Responses by College Students to a Dog and a Cat: Implications for Pet Therapy.” North American Journal of Psychology 10 (3).

Tennekes, Martijn. 2020. Tmaptools: Thematic Map Tools. https://CRAN.R-project.org/package=tmaptools.

Wickham, Hadley. 2011. “The Split-Apply-Combine Strategy for Data Analysis.” Journal of Statistical Software 40 (1): 1–29. http://www.jstatsoft.org/v40/i01/.

———. 2016. Ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York. https://ggplot2.tidyverse.org.

Wickham, Hadley, Romain François, Lionel Henry, and Kirill Müller. 2020. Dplyr: A Grammar of Data Manipulation. https://CRAN.R-project.org/package=dplyr.

Wickham, Hadley, and Lionel Henry. 2020. Tidyr: Tidy Messy Data. https://CRAN.R-project.org/package=tidyr.

Wilkins, David. 2019. Treemapify: Draw Treemaps in ’Ggplot2’. https://CRAN.R-project.org/package=treemapify.

Zhu, Hao. 2019. KableExtra: Construct Complex Table with ’Kable’ and Pipe Syntax. https://CRAN.R-project.org/package=kableExtra.